Reddit user creates benchmark to simulate personal AI use and detect downgrades

The benchmark evaluates AI model performance within individual workflows and flags regressions after updates. It aims to fill gaps in traditional benchmarks that don't reflect personal usage or post-benchmark downgrades.